I found a fix for accessing BCEAO data web site: repeat initial request twice while saving and sending back cookies The web site has also changed with new data set. The most noticeable change is the disappearance of dataset codes from the initial page. One must now read them on each individual dataset page html.
As a consequence, I moved building the category_tree to download
Two datasets are missing so I used an exception to make sure that download reaches the end even if some datasets fail downloading
(edit: reformat description)